A Brainy Twist to Image Analysis

Courtesy of DefiniensBromodeoxyuridine-stained tissue of mouse small intestine (original image, left; and Cellenger analysis, right). Cellenger has extracted whole crypts on a larger scale and distinguishes between mitotic and non-mitotic nuclei on a smaller scale.Propelled by the pharmaceutical industry, the high-content imaging field has experienced rapid growth, with several major new instrument releases in the past two years. High-throughput image analysis has lagged behind, however, because

Aileen Constans
Jun 6, 2004
<p></p>

Courtesy of Definiens

Bromodeoxyuridine-stained tissue of mouse small intestine (original image, left; and Cellenger analysis, right). Cellenger has extracted whole crypts on a larger scale and distinguishes between mitotic and non-mitotic nuclei on a smaller scale.

Propelled by the pharmaceutical industry, the high-content imaging field has experienced rapid growth, with several major new instrument releases in the past two years. High-throughput image analysis has lagged behind, however, because current software packages cannot keep pace with the huge amount of data these systems generate, says Martin Baatz, director of business development at Munich-based Definiens http://www.definiens.com. Definiens' new Cellenger software addresses this problem, providing fully automated quantitative analysis of biological images.

Cognition Network Technology, the platform on which Cellenger is based, was developed over the past nine years by Definiens' cofounder and Nobel laureate Gerd Binnig and his team. Initially written for the geoscience community, which used it to extract interesting...

COGNITIVE AWARENESS

In standard image-analysis software, regions of interest are identified based on feature similarities between neighboring pixels. While this approach works well when the features in the object of interest are relatively homogenous, it can fail in more complex situations, for example, the identification of diseased cells, organelles, or tissue structures. In contrast, Cognition Network Technology groups pixels into "information units" according to user-defined parameters called rule sets.

The software uses a knowledge-based approach in which the image objects gradually become more fine-tuned, allowing the user to ultimately extract structures of interest even from highly heterogeneous images. "You start out with something that is just a bunch of pixels, and you sort of tease in a step-by-step manner a little bit more information out of your image as you work your way to an end result. And this, I think, is a fairly intuitive way of trying to understand image processing," says Kurt Anderson of the Max Planck Institute of Molecular Cell Biology and Genetics, Dresden, Germany. Anderson currently uses Cellenger for several projects, including the analysis of cells during Drosophila development.

Cellenger has a range of applications, including diagnostics and electron microscopy, but Baatz says it is now mostly used for cell-based screening and histology/pathology studies. The preclinical safety group at Basel, Switzerland-based Novartis Pharmaceuticals currently uses the software to study cell proliferation in the crypts of mouse small intestine.

Previously, these studies required a user to screen 15,000 images manually, as existing software could not extract the crypts from the background. Manual analysis took up to seven months to compete; Cellenger reduced this time to two months. "[Manual detection] was very labor-intensive and boring. With Cellenger it was possible to automatically detect this crypt area," says Elke Persohn, head of Novartis' cellular imaging group.

GROWING DEMAND

Definiens offers two Cellenger products: Enterprise, a complete image-analysis package designed for end users; and Developer Studio, which allows users to design their own rule sets, or algorithms, for analyzing data. At least one user sees some room for improvement. Anderson says he would like to see "some sort of user-friendly items on the rule-set development side that would allow people writing their own rule set for the first time to do it more quickly and to understand what happens when [they] change a certain parameter" and what effect these changes have on the final analysis.

Anderson concedes, though, that this suggestion may be a "fine point" and that, overall, he has been impressed by the technology. Baatz says the company has received positive feedback from current users and potential customers. "I'm not aware of any commercially available product that even does close to what this does," says Anderson.

Definiens has formed partnerships with hardware vendors, including Evotec Technologies of Hamburg, Germany, and offers Cellenger to commercial customers for €35,000, with a 50% discount to academic laboratories. Baatz says the company plans to open a US subsidiary or headquarters in the near future to meet growing demand for its products.

- Aileen Constans

Interested in reading more?

Become a Member of

Receive full access to digital editions of The Scientist, as well as TS Digest, feature stories, more than 35 years of archives, and much more!
Already a member?